Skip to main content
Literature Notebook
Create free account

Policies

AI disclosure

We would rather be dull and accurate than fashionable and vague. Here is exactly where artificial intelligence does and does not appear in Literature Notebook.

Last updated September 17, 2026

Today: one opt-in assistant, on Studio

The Studio Assistant is the only AI feature in the app. It runs only when you open it and send a request, and only on the project, section or sources you choose. Nothing is sent because you opened a page, and no text is added to a draft unless you review the proposed change and apply it yourself. Every assistant reply names the material it was given.

On Notebook Free and Notebook Plus the assistant is not available at all. When an assistant request cannot be completed, we say so instead of returning invented text. Everything else in the product — the adaptive pages, comparisons, matrices and revision plans — is hand-authored framework and your own writing, not model output.

Which service and model an assistant request uses

Assistant requests are sent to the managed Lovable AI Gateway, which routes them to the configured model openai/gpt-5.6-sol. That is the active configuration current as of 17 September 2026; it is a statement of fact today, not a permanent commitment. If we change the service or the model, this page is updated with a new date.

The assistant runs on the server. The key that reaches the service is never sent to your browser, and your browser never contacts a model provider directly.

Exactly what is sent

Only your instruction plus the specific works, sections, passages or thread messages you added to this request are sent. Your account, library and files are not sent.

Your notebook is not searched or swept for context. The request carries the items listed above the composer in the Assistant panel — and only those — together with the internal instructions that keep the reply grounded and cited. Uploaded files are not sent.

Retention and training, stated honestly

We do not use your works, highlights, notes, sources, manuscripts or assistant threads to train or fine-tune models of our own, and we do not sell or share them for that purpose. That commitment is in the Terms of Service and the Privacy Policy, not only on this page.

What happens to a request after it leaves us is governed by the terms of Lovable AI Gateway and of the model provider it routes to. We have no contract with them that lets us promise zero retention or zero training on their side, so we do not make that promise. Treat an assistant request as material shared with a third-party processor, and do not send anything you cannot share on those terms.

Their currently published policies, for reference. These are the providers’ own documents, not an agreement we hold on your behalf, and they can change without notice:

Your threads and replies are stored in your own account, appear in your data export, and are deleted when you delete the thread or the account.

The rules we hold the assistant to

These rules apply to the Studio Assistant now, and to anything similar we add later. You can see how the assistant is billed on the pricing page.

  • It is opt-in per action. Nothing gets sent anywhere because you opened a page.
  • Generated passages are marked as generated and arrive as a proposal you accept or reject, never silently merged into your draft.
  • Only the project material you select is sent, and a request that reaches for anything outside your access is refused by the server.
  • It assists analysis and revision — it does not write your work for you and present it as yours.
  • The service and model in use are named on this page with the date they were confirmed, and we describe upstream retention only as far as we can honestly stand behind it.
  • If we change the provider or the purpose, this page and the privacy policy are updated first.

A note on how the product was built

AI-assisted development tools were used to help build this application, in the same way an editor or compiler is used. That has no bearing on your content, which stays private to your account.